©2026 Business Next Media Corp. All Rights Reserved. No.102, Guangfu S. Rd., Da'an Dist., Taipei City 106, Taiwan
Taiwan-U.S. Collaboration Powers the Silicon Photonics Ecosystem
With the explosive growth of artificial intelligence, the demand for high-performance computing has reached unprecedented levels. However, current computing infrastructure faces a serious data-transfer bottleneck. While chip speeds continue to improve, the data transfer efficiency between chips lags behind due to physical limitations. This issue causes severe data congestion within AI data centers, restricting AI model performance and resulting in significant energy loss through excess heat.
To address this critical issue, the industry is turning to Optical Interconnect (Optical IO) technology, which uses light instead of electrical signals to transfer data. Among these technologies, Co-Packaged Optics (CPO) has received the most attention. CPO integrates optical components directly onto electronic chips to enhance efficiency.
However, existing CPO solutions still face major challenges. These include larger sizes, difficulty integrating light sources, low manufacturing yields, high costs, and heat dissipation issues. During Meet Global’s AI Online Happy Hour, Taiwan-based startup Rayleigh Vision and U.S.-based silicon photonics unicorn Ayar Labs shared insights into potential breakthroughs for overcoming these core technical obstacles.
Rayleigh Vision: AI-driven micro-emitters with near-perfect manufacturing yield
"Copper-based interconnect technology has already hit its limits," said Jr-Hau He, founder of Rayleigh Vision. "We need a new solution."
Based in Hsinchu, Taiwan, Rayleigh Vision specializes in AI-assisted design and manufacturing of MicroLED technology. Using a unique mass-transfer process, the company precisely places thousands of tiny optical components onto silicon modules with micron-level accuracy. This technology could enable the creation of micro-emitter arrays on silicon chips, overcoming the fundamental limitation of chips being unable to produce optical signals.
A significant breakthrough by Rayleigh Vision is the development of the world's first AI-driven production line. This innovation not only enables precise placement of components but also has the capability to repair individual emitters. Thanks to this AI-enabled precision repair technology, Rayleigh Vision’s modules can achieve manufacturing yields approaching 100%, even when containing hundreds of microchips. This effectively addresses the low yield and high cost issues currently seen with conventional CPO manufacturing processes.
“Our technology can deliver unprecedented performance,” emphasized Jr-Hau He. “With a bandwidth of 2.5 Tbps per square millimeter, high-performance computing becomes even more attainable.”
Rayleigh Vision targets hyperscale data centers requiring high bandwidth and low power consumption, as well as AI factory markets needing low-latency integration of computing, memory, and optical components through 3D stacking. Positioning themselves as industry enablers, Rayleigh Vision aims to bridge the gap between emitters and integration, allowing chip manufacturers to concentrate on their core computing functions.
The company has already begun shipping its products, targeting $100 million in revenue by 2028. In response to geopolitical factors and the localization of supply chains, Rayleigh Vision plans to open a U.S. office by the end of this year. Additionally, the company is considering licensing its optical emitter design and mass-transfer technology to partners in the United States.
Ayar Labs: Optical IO chipsets and building a broad ecosystem
Ayar Labs, founded in 2015, is a Silicon Valley unicorn with a valuation exceeding $1 billion. The company has attracted strategic investments from major players such as AMD, Intel, and NVIDIA. Ayar Labs focuses on optical interconnect solutions, particularly addressing the data-transfer bottlenecks in high-performance computing. They see copper-wire technology as having reached its limit, making optical interconnect solutions essential.
“Future computing demands will only continue to grow,” said Terry Thorn, Vice President of Business Development at Ayar Labs. “Optical IO is the logical solution to the data-transfer bottleneck.”
The company's core products include the TeraPhi optical IO chipset, which performs electro-optical conversion and supports multi-chip packaging using standard CMOS processes, and the Supernova product line, which provides a robust external light source. Ayar Labs actively supports industry-standard interfaces such as UCIE, helping accelerate the widespread adoption of optical technologies.
Terry Thorn acknowledged that achieving large-scale commercialization faces significant non-technical challenges. These include establishing a comprehensive high-volume manufacturing ecosystem with wafer fabs, OSAT companies, and testing partners. Additionally, promoting new standards such as UCIE and ensuring the quality and reliability of integrated optical components are crucial. Ayar Labs is currently expanding its presence in Hsinchu, Taiwan, actively recruiting engineers and R&D talent to leverage Taiwan’s complete ecosystem of manufacturing, assembly, and testing resources.
Industry Challenges and Taiwan-U.S. Ecosystem Collaboration
Both Rayleigh Vision and Ayar Labs emphasized that widespread adoption of optical interconnect technology requires more than just technological breakthroughs—it also depends heavily on a mature and collaborative industrial ecosystem. Hsinchu, Taiwan, is home to world-class IC design and manufacturing infrastructure, making it a critical location for achieving large-scale production of optical interconnect technology. Ayar Labs’s expansion and talent recruitment in Taiwan, along with Rayleigh Vision’s establishment of a U.S. office and potential technology licensing, demonstrate the complementary nature and strong potential for collaboration between the U.S. and Taiwan in optical interconnect technology.
Silicon photonics and optical interconnect technology are becoming essential solutions to address data-transfer bottlenecks in the AI era. Through technological innovation, enhanced manufacturing capabilities, and close ecosystem collaboration, both companies are paving the way toward a broader, faster, and more efficient optical highway for the future of AI computing.
©2026 Business Next Media Corp. All Rights Reserved. No.102, Guangfu S. Rd., Da'an Dist., Taipei City 106, Taiwan